Probability Brainteasers for Quant Interviews: Dice, Coins, Cards, Urns

Probability Brainteasers for Quant Interviews: Dice, Coins, Cards, Urns

Quant interview rounds at Jane Street, Citadel, Two Sigma, Optiver, SIG, and most hedge funds rely heavily on probability brainteasers. The questions vary in surface form — coin flips, dice rolls, decks of cards, urns full of balls — but they test a small set of mental tools: expected-value linearity, conditional probability, recursive reasoning, indicator variables, and symmetry arguments. Memorizing answers won’t help; interviewers vary problems specifically to detect rote prep. This guide covers the canonical problems, the techniques behind them, and how to think about them under interview pressure.

Foundational Tools

Linearity of Expectation

The most powerful tool. E[X + Y] = E[X] + E[Y], even when X and Y are dependent. This lets you compute expected values of complex events by decomposing them into simpler indicator variables and summing.

Example: “Roll a fair die 100 times. Expected number of times you see two consecutive 6s?” Define indicator X_i = 1 if rolls i and i+1 are both 6, else 0. Then total = sum of 99 indicators, each with E[X_i] = 1/36. Answer: 99/36 ≈ 2.75.

Recursion / Conditioning on First Step

Many problems yield to setting up a recursive equation by conditioning on the first event. “What’s the expected number of coin flips until two heads in a row?” Let E be the answer. With probability 1/2, first flip is tails — we waste 1 flip and start over: E = 1 + E. With probability 1/2, first flip is heads. Then with probability 1/2, second flip is heads (done in 2 flips); with probability 1/2 it’s tails (wasted 2 flips, start over). So E = (1/2)(1 + E) + (1/2)((1/2)(2) + (1/2)(2 + E)) → solve to get E = 6.

Indicator Variables

For “expected number of X” problems, define an indicator for each potential occurrence and sum. Often easier than computing the full distribution.

Symmetry

If a problem has a natural symmetry, exploit it. “Two players draw cards from a shuffled deck without replacement. Probability the first ace appears in player A’s draws?” By symmetry over the four ace positions, the answer is 13/52 / (13/52 + 13/52) wait that’s not right — actually use direct argument: each ace is equally likely to be at any of the 52 positions. Player A’s draws are positions 1, 3, 5, …, 51 (26 of them). So probability the first ace is in A’s draws = 26/52 = 1/2. Symmetry argument works.

Coin Flip Problems

Expected flips until two heads in a row

Answer: 6. Recursive setup as above.

Expected flips until heads-then-tails (HT)

Answer: 4. Setup: let E be expected flips until HT. First we need an H (expected 2 flips for the first head, since geometric with p=1/2 has mean 2). Once we have an H, we need a T immediately (expected 2 flips for the next T-or-H sequence’s resolution — if T, done; if H, we’re still “post-H” and need another flip). E[flips after first H to get T] = 2. So E = 2 + 2 = 4.

Compare HH vs HT

Common follow-up: “I’ll flip a coin until I see HH; you flip until you see HT. Whose expected number of flips is shorter?” Answer: HT (4 flips) is shorter than HH (6 flips). The asymmetry comes from the “reset” structure of HH — if you see HT after the first H, you have to start over from scratch, while in HT every H sets up a one-flip-to-resolve situation.

Probability of HH appearing before HT in a random sequence

Trickier. Both must contain an H first. After the first H, the next flip is either T (giving HT, T wins) or H (giving HH, H wins). So probability HH first = 1/2.

Two players race to flips: I get heads first, you get tails first — probability I win?

If we alternate flips and I go first: P(I win) = P(my first flip is H) + P(my first is T, you get H, my second is T, you get H, …) = 1/2 + (1/2)(1/2)(1/2)… pattern. Actually clearer: I win on flip 1 with prob 1/2; on flip 3 with prob (1/2)(1/2)(1/2) = 1/8; etc. Geometric series: 1/2 + 1/8 + 1/32 + … = (1/2) / (1 – 1/4) = 2/3.

Dice Problems

Expected rolls of a fair die until you see a 6

Geometric distribution with p = 1/6, so expected rolls = 1/p = 6.

Expected rolls until you see all six faces

Coupon collector. Expected rolls = 6(1 + 1/2 + 1/3 + 1/4 + 1/5 + 1/6) = 6 × 49/20 = 14.7.

Roll three dice. Probability all three sum to 18?

Only one way: (6,6,6). Probability = 1/216.

Roll a die. If you get N, you roll N more dice and sum the result. Expected total?

First roll has E = 3.5. Conditional on first roll = N, expected sum of next N dice = 3.5N. So total expected = 3.5 + E[N] × 3.5 = 3.5 + 3.5 × 3.5 = 3.5 + 12.25 = 15.75.

St. Petersburg paradox variant

“Roll a die. If you get N, you win 2^N dollars. What would you pay to play?” Expected value = (1/6)(2 + 4 + 8 + 16 + 32 + 64) = 126/6 = $21. (Real St. Petersburg is unbounded; this version isn’t.)

Card / Deck Problems

Expected position of the first ace in a shuffled deck

Use indicator variables. Let X_i = 1 if the i-th card is before all four aces. Then position of first ace = 1 + sum of X_i for non-ace cards. Each non-ace has probability 1/5 of being before all four aces (by symmetry: 5 positions for any group of 4 aces + 1 specific non-ace). Wait, this gets complicated; the cleanest way: position of first ace = (53 – 4) / (4 + 1) + 1 = 48/5 + 1 = 10.6. (The 5 spaces argument: 4 aces divide the 48 non-aces into 5 equal-expectation gaps.)

You draw cards one at a time without replacement until you get an ace. What’s the probability the previous card was a king?

Surprising answer: 4/51. Argument: pair each ace with the card immediately preceding it. There are 52 cards and 52 possible “previous” positions if we treat the deck as a circle. By symmetry, the probability that any specific card is immediately before a specific ace is uniform; there are 4 aces and 4 kings, so probability = 4×4 / (52 × …) → cleaner: think of the 52! arrangements and count those where a king immediately precedes the first ace. Result: 4/51.

Two cards drawn without replacement. Probability both are red?

(26/52)(25/51) = (1/2)(25/51) = 25/102 ≈ 24.5%.

Five cards dealt; what’s the probability of at least one pair?

1 minus probability all distinct ranks. P(all distinct) = (52/52)(48/51)(44/50)(40/49)(36/48) ≈ 0.507. So P(at least one pair) ≈ 49.3%.

Urn / Ball Problems

Urn has W white and B black balls. Draw without replacement; expected number of color changes in the sequence?

For each adjacent pair of positions, probability of a color change = (W × B) / [(W+B)(W+B-1)] × 2 (for white-then-black or black-then-white). With W+B-1 adjacent pairs: expected changes = 2WB / (W+B). For W=B=N, this is N. For W=10, B=10, it’s 10 color changes on average.

Two urns, urn A has 3 white and 1 black, urn B has 1 white and 3 black. Pick urn at random, draw a ball, it’s white. Probability you drew from urn A?

Bayes: P(A | W) = P(W | A)P(A) / P(W) = (3/4)(1/2) / [(3/4)(1/2) + (1/4)(1/2)] = (3/8) / (4/8) = 3/4.

Polya’s urn

Urn starts with 1 white and 1 black. Each draw: observe color, return ball, add another of same color. After N draws, expected fraction white?

Surprising: 1/2 always. The process is exchangeable; the marginal probability of each draw being white is 1/2 regardless of N. The variance is high — eventually one color dominates — but the expected fraction stays at 1/2.

Random Walk Problems

Random walk on a number line, starting at 0. Probability you reach +N before -M?

For a fair walk: P = M / (N + M). Symmetric and depends only on the relative starting distances.

Expected time to reach +N starting from 0 (fair walk)?

For a walk that ends at +N starting from 0, with no lower bound: expected time = ∞ (walk can go arbitrarily negative). With absorbing boundaries at +N and -M: expected time = NM (a famous result).

Drunk in a one-dimensional walk: probability of returning to start eventually?

1 (Pólya’s theorem for 1D and 2D). In 3D, probability of eventual return is approximately 0.34.

Two-Player Game Problems

Two players take turns flipping a coin; first to flip heads wins. I go first. P(I win)?

Geometric: 1/2 + (1/2)²(1/2) + (1/2)⁴(1/2) + … = (1/2) / (1 – 1/4) = 2/3.

Each draws a card; whoever gets the higher card wins. Tie-breaking on suit (spade > heart > diamond > club)?

By symmetry plus tie-breaker, no ties exist. P(I win) = 1/2 by symmetry.

Multi-stage betting: I bet 1 unit, you bet 2, I bet 4 (doubling). Continue until one runs out. Long-run dynamics?

This is the gambler’s ruin problem with doubling stakes. Discuss when bankrolls are bounded vs. unbounded; in either case the long-run analysis depends on starting capital.

How to Approach Brainteasers in Interviews

Step 1: Restate the problem clearly

Confirm understanding before computing. Rushed misinterpretations cost candidates more than slow correct answers. Say: “So we’re flipping a fair coin until we see two heads consecutively. We want the expected number of flips. Let me set up a recursion…”

Step 2: Identify the technique

Match the problem to one of the foundational tools: linearity, recursion, indicators, conditioning, symmetry. Verbalize: “This looks like a linearity-of-expectation problem; let me define indicators…”

Step 3: Compute carefully out loud

Talk through the computation. Interviewers want to see your reasoning, not just the final answer. If you make an arithmetic mistake, they’ll catch it before you commit. If you have the right approach but wrong answer, the interview survives.

Step 4: Sanity-check the answer

Does the magnitude make sense? “6 flips for HH feels right because it’s longer than expected first-head (2) but not absurdly so.” If the answer is dramatically counterintuitive, double-check.

Step 5: Be ready for follow-ups

Interviewers often extend brainteasers: “Now what if the coin is unfair, P(H) = p? What’s the expected number?” or “What if we want HHH instead of HH?” Practiced candidates handle the extension fluently because they understand the underlying mechanics.

Practice Strategy

Weeks 1–2: Work through the foundational tools section. Practice 30 problems mixing categories. Focus on getting the technique right, not speed.

Weeks 3–4: Increase to 5 problems per day, timed (5 minutes each). Force yourself to verbalize approach before computing.

Weeks 5+: Mock interview practice. Have someone present problems you haven’t seen and grade your reasoning, not just answers.

Resources: Heard on the Street by Crack (canonical), A Practical Guide to Quantitative Finance Interviews by Zhou (more comprehensive), Quant Job Interview Questions and Answers by Mark Joshi (more derivatives-heavy).

Frequently Asked Questions

See also: Breaking Into Quant Finance and Wall Street: 2026 Guide • Jane Street Interview Guide • Citadel Securities Interview Guide (coming soon)